Security
Trust is earned with specifics, not badges. Here is exactly how we protect the service today, what we do not claim, and how to tell us when we got something wrong. We review this page periodically and whenever these controls change.
The short version
Minimal data, two-step verification, sessions you can see and revoke, an audit log for every workspace, secrets encrypted with rotating keys, and a codebase anyone can read.
Compliance status, plainly: zerosoft does not hold a SOC 2 report or an ISO 27001 certificate, and no independent audit or penetration test has been performed yet. Our controls are designed to meet the SOC 2 Trust Services Criteria and ISO/IEC 27001:2022 Annex A, and the policies and control matrix are public in our repository (docs/security). When an audit starts or finishes, this page will say so.
Signing in
There are no zerosoft passwords. You sign in with Google (verified email addresses only) or with a one-time code sent to your email.
You can turn on two-step verification with any authenticator app (TOTP), with ten single-use recovery codes. Workspace owners can require it for every member: someone without it cannot open that workspace until they turn it on.
Every sign-in is a session on our side. In Settings › Security you see your devices, when each was last active, and you can sign out one or all of them; signing out ends that session on the server too. Sessions expire after 30 days without use.
Access control and audit log
Each workspace has owners, admins, members and guests, access per app and per record. Every page and API checks these rules on the server.
Owners and admins see a security audit log: sign-ins, two-step changes, sessions closed, API keys created or revoked, exports, deletions and member changes, with time and IP. It can be downloaded as CSV and is kept for one year.
API keys
Programmatic access uses API keys with the zs_ prefix. Keys are generated from a cryptographically secure random source, shown to you exactly once, stored only as SHA-256 hashes, scoped to the apps you choose, and revocable instantly. A member's keys stop working when they leave the workspace.
Each key is rate-limited to 100 requests per minute.
Link passwords
Password-protected short links store their passwords as scrypt hashes with a random salt per link. Once a visitor unlocks a link, a signed cookie remembers it for up to 7 days, and attempts are rate-limited per IP and per link.
Encryption
In transit: all traffic uses HTTPS with modern TLS, plain HTTP is redirected, and browsers are told to use HTTPS only (HSTS). Connections to our database and to our own services also use TLS.
At rest: our infrastructure providers encrypt stored data and backups. On top of that we encrypt secrets ourselves before they reach the database (OAuth tokens, third-party API keys, webhook secrets, connected-account settings, credentials and two-step secrets) with AES-256-GCM, a separate key for every value, and master keys that we can rotate. Private files such as CVs and signed PDFs are encrypted before upload.
End to end: locked notes are encrypted in your browser and we cannot read them. We do not encrypt everything end to end, because search, AI features, integrations, automations and real-time collaboration need the server to read the content. No security standard requires it; we use it where the server does not need to read the data.
Uploaded files
Files attached in the apps are stored at URLs with long random names. They are hard to guess but not private: anyone with the URL can open the file. File types that browsers would execute (HTML, SVG, XML and JavaScript) are rejected. Private files (CVs, signed documents, receipts) never get a public URL.
Data minimization
Click analytics are cookie-less and fingerprint-less: referrer domain, country and device are derived from request headers, and the IP address is never stored with the click.
IP addresses are used briefly for rate limiting and are recorded with sign-ins and sessions so you can recognize your devices; those records are deleted after one year (audit log) and 90 days after a session ends.
Abuse protection
Public endpoints, sign-in codes and two-step codes are rate-limited. Outgoing requests to URLs you configure, such as webhooks, are restricted to public internet hosts with timeouts. We may disable short links used for phishing, malware or spam; report abuse to abuse@zerosoft.ai.
How we build
Every change goes through a pull request and automated checks before it ships: type checks, unit and end-to-end tests, a production build, static analysis with CodeQL, secret scanning and a dependency audit. Dependabot watches our dependencies, and security-relevant changes get an independent review.
Infrastructure and subprocessors
The cloud runs on Vercel (hosting, functions, files), Turso (database) and Railway (WhatsApp gateway), sends email with Resend, and uses AI model providers only for AI features you use. The full list, with what each one processes, is on the subprocessors page (/subprocessors). Their certifications are theirs, not ours. Access to production is limited to the people who need it and protected with two-step verification.
Your data
In Settings › Profile you can download everything in your account and workspace as a ZIP, and delete your account or your workspace. Deletion happens 30 days after you ask (you can cancel until then) and covers database records, uploaded files and connected WhatsApp numbers; copies in provider backups expire within their retention window.
Customers who need a data processing agreement can request ours at privacy@zerosoft.ai.
Open source as an audit
The code of the products is public. You do not have to trust this page: you can read the code, run it yourself and check every claim. Reports from people who do exactly that are our favorite kind of email.
Reporting a vulnerability
Email security@zerosoft.ai with a description, steps to reproduce and the impact. Please do not open public issues for security problems. We aim to acknowledge within 72 hours, keep you updated, and credit you when the fix ships if you want. We do not run a paid bug bounty.
Test only against your own accounts and data, do not degrade the service, send spam or use social engineering, and give us a reasonable window before publishing. If you act in good faith, we will not pursue legal action against you.
If something goes wrong
If we learn of a security incident that affects your data, we investigate and contain it, notify affected account holders without undue delay so they can meet their own 72-hour duties, and notify data protection authorities when the law requires it. Service incidents are posted on /status.